opamp bypass caps with regulated PS

[Russ White]

I am designing a preamp with a regulated PS.

The PS will be +/- 12V with 7812 and 7912 regulators after 2 1500uf filter caps per rail.

The preamp is based on the LT1115 opamp and the LT1010 buffer. (I have already built a couple opa637/buf634 circuits, so I am trying something new).

I usually use unregulated supplies and put 10uf tantalums with .1uf ceramics at the supply pins.

I have two questions:

1) When using a regulated supply are the bypass caps as critical? I was going to use them anyway, and I had chosen to use just the .1 uf high quality ceramics alone. Would that be OK?

2) Should I add a large capacitor after the regulator to help with transient response, say one more 1500uf pnanasonic FC per rail.

Am I on the right track?

Thanks for your help.

[Stocker]

I think the results people hear with the snubberized power supply concept from IIRC CarlosFM is a good indicator that a combination of big and small caps even after regulators is a good idea.

[macboy]

1) When using a regulated supply are the bypass caps as critical? I was going to use them anyway, and I had chosen to use just the .1 uf high quality ceramics alone. Would that be OK?

Yes they are just as critical. They are not there to filter the power supply, they are there to provide a low impedance power supply to the load (opamps). The inductance of the power supply traces/wires leading to the opamps would otherwise cause problems.

2) Should I add a large capacitor after the regulator to help with transient response, say one more 1500uf pnanasonic FC per rail.

No. Most regulators do not like large capacitances on their output. Low dropout voltage (LDO) type regulators are especially picky about this, but even conventional ones aren't too happy about it. Let the regulator do its job of actively providing a low impedance voltage source. Make sure to have good filtering (and bypassing!) on the input side though. You can generally use a electrolytic up to 100 uF without issues, and a small ceramic can usually help reduce output noise too.

[Russ White]

Thanks man, I have been using the regulators with 1000uf on the output! It has not seemed to cause any problem, but I will remove them and see how things go.

Thanks again. I really appreciate the input.

[peranders]

Don't forget that there are levels of everything.

To little decoupling = oscillations, unstability, noise etc.

Enough = things work good

More = some drawbacks of some parameters and some will be better

Overkill = Maybe more drawbacks. Not certain that the PS will work good at all.

[mauropenasa]

lead( if you want ) a 0.1uF or 0.22uF Ceramic multilayer or MKT on the OPamp ( as have done in my schemes ) among +Vcc and -Vcc. Works as says macboy, In more improve the CMRR of the chips, without problems of ground loop.
For the rest, full sharing with macboy and peranders

Ciao

Mauro

[janneman]

Originally posted by macboy

Yes they are just as critical. They are not there to filter the power supply, they are there to provide a low impedance power supply to the load (opamps). The inductance of the power supply traces/wires leading to the opamps would otherwise cause problems.
[snip]

If you are realy serious about this, use remote sensing of the supply AT THE OPAMP. That gives you the lowest impedance. Using a low impedance supply and then solving the connection wiring witha suboptimal cap at the opamp is very sub-optimal, especially at mid- to high frequencies.

Originally posted by macboy
[snip][No. Most regulators do not like large capacitances on their output. Low dropout voltage (LDO) type regulators are especially picky about this, but even conventional ones aren't too happy about it. Let the regulator do its job of actively providing a low impedance voltage source. Make sure to have good filtering (and bypassing!) on the input side though. You can generally use a electrolytic up to 100 uF without issues, and a small ceramic can usually help reduce output noise too.

No. Regulators (all) NEED cap at the output for stability. There is a minimum cap, but no maximum, although with very large caps you get to the point of diminishing returns. Especially for lf transients, there is nothing wrong with a couple of 1000 uF at the supply output, but that's probably the max that will improve the results. LDO reg are often designed to get away with low cap at the output, 1uF or even less. That is done for commercial reasons, but more than that doesn't harm at all and improves transient response.

Jan Didden

[Russ White]

Thanks Jan,

I have re-read the data sheet for the LM7812 and LM7912 which I am using, and I agree with you that all I see is that capacitance at the output is a good thing.

If you are realy serious about this, use remote sensing of the supply AT THE OPAMP. That gives you the lowest impedance. Using a low impedance supply and then solving the connection wiring witha suboptimal cap at the opamp is very sub-optimal, especially at mid- to high frequencies.

I wish I knew what you meant by this. I still have a lot to learn. Could you give me some guidance. Or point me in the right direction for research?

Thanks.

BTW the PS is built and running fine. See my GC preamp thread.

[janneman]

Russ,

The original so called 'jung' regulators (article series in Audio Amateur 1995 by Walt Jung, yours truly and Gary Galo) were specifically laid out to cope with this problem.
As you know, in a reg, a sample of the output (usually via a two-resistor divider) is fed back to the inverting opamp input to close the regulation loop. What that means is that the thing regulates the voltage AT THE DIVIDER! In the jung type regs, that divider was put at the opamp or circuit -to-be-supplied, to make sure that you get the cleanest supply where it matters, at the load, hence the name remote sensing.
Now, there is more to it, because it also needs special care in the grounding, so you need to design this in right from the start.
Since you have already the supply running, you can optimise it by keeping the hot leads to the load as short as possible. Also, make sure that you don't use a ground lead that carries more than one signal, like both the supply return and the signal ground, all that should be run separately to a common star ground that also receives the common of the input circuits, like xformer mid point (if used) and the filter capacitors.

Jan Didden

[macboy]

No. Regulators (all) NEED cap at the output for stability. There is a minimum cap, but no maximum, although with very large caps you get to the point of diminishing returns. Especially for lf transients, there is nothing wrong with a couple of 1000 uF at the supply output, but that's probably the max that will improve the results. LDO reg are often designed to get away with low cap at the output, 1uF or even less. That is done for commercial reasons, but more than that doesn't harm at all and improves transient response.

I beg to differ. There are many very different regs out there and you are claiming that all are alike and behave similarly. LDO regulators can easilly go into oscillations with as little as a couple 100 uF of low ESR caps at the output. You can put more capacitance, but you need to either use a cap with moderate to high ESR or put a resistor in series with the cap to effectively increase ESR (the resistance damps the resonant circuit preventing oscillations). Non-LDO regs like 78xx are much more stable, but can be made to oscillate with the wrong kind of load (including the output capacitance) at the output. But since there are dozens of different 78xx's from many different manufacturers, I believe that no generalizations can be made. Blindly implementing "no maximum", or "more ...doesn't harm at all " may be fine for brand-X, but may send brand-Y into a tizzy.

I'd also point out that the position of these caps can make a difference too. You can often get away with larger caps near the load (far from the regs) than you could at the regs, due to the series impedance of the power supply tracks. And near the load is where they will do the most good, by providing low impedance where it is needed. Then again, this big capacitance combined with thin, inductive power supply traces, can cause oscillations. So which is better? Only a well-equipped electrical engineer can do the required analysis to determine that. A lot of DIY'ers love to throw capacitance at power supplies without understanding that more isn't always better.

If you are realy serious about this, use remote sensing of the supply AT THE OPAMP. That gives you the lowest impedance. Using a low impedance supply and then solving the connection wiring witha suboptimal cap at the opamp is very sub-optimal, especially at mid- to high frequencies.

This can cause all kinds of problems, since you are putting the inductance of the power supply traces/wires (between the reg and the load/opamp) into the feedback loop of the regulator. This inductance, combined with output capacitance creates a L-C resonant circuit. This can cause problems, including HF oscillations or much worse. This can be made to work, but it isn't as simple as you've implied it to be. I requires a pretty in-depth engineering analysis to ensure stability, and in most cases, you'll need to add some compensating elements to basically reduce the speed of the regulator so that it won't cause the L-C resonant circuit to oscillate. So remote sensing can, and does, improve DC regulation (the accuracy of the DC voltage) but generally reduces transient response. Arguably, in audio, transient response is more important than DC accuracy. It is important to note the difference between a power supply with excellent regulation (little dipping of voltage when under increased load) and a clean power supply (low AC noise).